Skip to content

[AMDGPU] Newly added test modified for recent SGPR use change #116427

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Nov 15, 2024

Conversation

JanekvO
Copy link
Contributor

@JanekvO JanekvO commented Nov 15, 2024

Mistimed rebase for #112251 which added new tests which did not consider the changes introduced in #112403 yet

@JanekvO JanekvO requested review from arsenm and shiltian November 15, 2024 19:39
@nico nico merged commit 9a5e5e2 into llvm:main Nov 15, 2024
5 of 9 checks passed
@llvmbot
Copy link
Member

llvmbot commented Nov 16, 2024

@llvm/pr-subscribers-backend-amdgpu

Author: Janek van Oirschot (JanekvO)

Changes

Mistimed rebase for #112251 which added new tests which did not consider the changes introduced in #112403 yet


Full diff: https://github.com/llvm/llvm-project/pull/116427.diff

1 Files Affected:

  • (modified) llvm/test/CodeGen/AMDGPU/function-resource-usage.ll (+1-1)
diff --git a/llvm/test/CodeGen/AMDGPU/function-resource-usage.ll b/llvm/test/CodeGen/AMDGPU/function-resource-usage.ll
index 278a2a0170ff33..22257d3eba7d63 100644
--- a/llvm/test/CodeGen/AMDGPU/function-resource-usage.ll
+++ b/llvm/test/CodeGen/AMDGPU/function-resource-usage.ll
@@ -594,7 +594,7 @@ define amdgpu_kernel void @usage_multi_stage_recurse_noattrs(i32 %n) #0 {
 ; GCN-LABEL: {{^}}multi_call_with_multi_stage_recurse:
 ; GCN:  .set multi_call_with_multi_stage_recurse.num_vgpr, max(41, use_stack0.num_vgpr, use_stack1.num_vgpr, multi_stage_recurse1.num_vgpr)
 ; GCN:  .set multi_call_with_multi_stage_recurse.num_agpr, max(0, use_stack0.num_agpr, use_stack1.num_agpr, multi_stage_recurse1.num_agpr)
-; GCN:  .set multi_call_with_multi_stage_recurse.numbered_sgpr, max(43, use_stack0.numbered_sgpr, use_stack1.numbered_sgpr, multi_stage_recurse1.numbered_sgpr)
+; GCN:  .set multi_call_with_multi_stage_recurse.numbered_sgpr, max(45, use_stack0.numbered_sgpr, use_stack1.numbered_sgpr, multi_stage_recurse1.numbered_sgpr)
 ; GCN:  .set multi_call_with_multi_stage_recurse.private_seg_size, 0+(max(use_stack0.private_seg_size, use_stack1.private_seg_size, multi_stage_recurse1.private_seg_size))
 ; GCN:  .set multi_call_with_multi_stage_recurse.uses_vcc, or(1, use_stack0.uses_vcc, use_stack1.uses_vcc, multi_stage_recurse1.uses_vcc)
 ; GCN:  .set multi_call_with_multi_stage_recurse.uses_flat_scratch, or(1, use_stack0.uses_flat_scratch, use_stack1.uses_flat_scratch, multi_stage_recurse1.uses_flat_scratch)

abidh pushed a commit to abidh/llvm-project that referenced this pull request Feb 4, 2025
…ass (llvm#102913)

Converts AMDGPUResourceUsageAnalysis pass from Module to MachineFunction
pass. Moves function resource info propagation to to MC layer (through
helpers in AMDGPUMCResourceInfo) by generating MCExprs for every
function resource which the emitters have been prepped for.

Fixes llvm#64863

[AMDGPU] Fix stack size metadata for functions with direct and indirect calls (llvm#110828)

When a function has an external call, it should still use the stack
sizes of direct, known, calls to calculate its own stack size

[AMDGPU] Fix resource usage information for unnamed functions (llvm#115320)

Resource usage information would try to overwrite unnamed functions if
there are multiple within the same compilation unit. This aims to either
use the `MCSymbol` assigned to the unnamed function (i.e.,
`CurrentFnSym`), or, rematerialize the `MCSymbol` for the unnamed
function.

Reapply [AMDGPU] Avoid resource propagation for recursion through multiple functions (llvm#112251)

I was wrong last patch. I viewed the `Visited` set purely as a possible
recursion deterrent where functions calling a callee multiple times are
handled elsewhere. This doesn't consider cases where a function is
called multiple times by different callers still part of the same call
graph. New test shows the aforementioned case.

Reapplies llvm#111004, fixes llvm#115562.

[AMDGPU] Newly added test modified for recent SGPR use change (llvm#116427)

Mistimed rebase for llvm#112251 which added new tests which did not consider
the changes introduced in llvm#112403 yet

Change-Id: I4dfe6a1b679137e080a6d2b44016347ea704b014
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants